Jamal Williams¹, William Narhi-Martinez², Anna Nobre²; ¹University of California, Santa Cruz, ²Yale University

As you enter the grocery store and check your list, you briefly hold several items in mind before deciding which one to grab first. In this instance, while much of the previous work would suggest that attention exerts the same influence over all items currently maintained in visual working memory (VWM), these findings are often based on simultaneous encoding. Yet, in the real world, representations are often encoded sequentially, thereby creating a “selection latency” between when an item is encoded and when it is prioritized by attention. Across several experiments, we demonstrate that this selection latency has a striking influence over the efficacy of attention, even when all representations are maintained with equal fidelity prior to being selected. In the current work, participants encode four real-world objects across two sequentially presented arrays: two objects appear in an initial “leading” array with the remaining two items presented in the “trailing” array. On half of all trials, a 100% valid retrocue indicates whether the tested item is drawn from the leading or trailing array; on the remaining "neutral" trials, the test item is selected at random. After a delay, a single test object is presented centrally and participants report its color from memory using a continuous color wheel. On neutral trials, memory error is effectively identical for items from the leading and trailing arrays. Yet, once an item is cued, we find a clear temporal asymmetry: despite all cued items showing reduced error compared to neutral items, those encoded more recently (ie, trailing items) show a profound benefit of attention and exhibit significantly reduced error compared to items from the leading array. This asymmetry persists across multiple control conditions, indicating that selection latency plays a critical role in shaping the efficacy with which attention operates over VWM representations.